Intellectual merit. The chemical industry is a vital sector of the US economy. Increasingly faced with the requirements of safety and profitability, chemical process operation is relying extensively on automated control systems, involving a large number of sensors. The reliance on sensors, however, tends to increase vulnerability of the process to sensor malfunctions (e.g., sensor failure, intermittent sensor data losses, biased measurements, etc.,), leading to the failure of the control system and potentially causing a host of economic, environmental, and safety problems that can seriously degrade the operating efficiency of the process. Management of abnormal situations resulting from sensor malfunctions is a challenge in the chemical industry since abnormal situations account for $10 billion in annual lost revenue in the US alone. The objective of this project is to develop a general and practical framework for handling sensor malfunctions in feedback control of chemical processes by explicitly dealing with sensor data losses and failures in the control system design and implementation. Nonlinear and predictive control theory will be used to produce practically-implementable, feedback control systems that account explicitly for the occurrence of sensor faults and enforce the desired stability, performance and robustness specifications in the closed-loop system. Hybrid systems and control theory will subsequently be used to: a) model and analyze sensor failure situations, and b) construct novel supervisory control schemes that ensure the timely and coordinated response of the local control systems in the process, in a way that achieves fault recovery and minimizes performance deterioration. The motivation is provided by: a) the common occurrence of sensor malfunctions in chemical process operation, b) the abundance of complex dynamics in chemical processes due to process nonlinearities, model uncertainties and constraints, c) the lack of practical control strategies for nonlinear chemical processes that can deal explicitly and simultaneously with complex dynamics, sensor data losses and sensor failures, d) advances in communication and computation technologies, and e) the continuing need to improve chemical process operation, reduce product variability, improve energy efficiency and minimize environmental and safety hazards. Specifically, the research will focus on: 1. Analysis and design of control and estimation systems subject to sensor data losses; both the state and output feedback control problems will be studied. 2. Design of integrated fault-tolerant control and estimation systems subject to complete sensor failures; both the sensor fault-detection and identification problem and the problem of sensor fault-induced control reconfiguration will be studied. 3. Control of multiple interconnected units subject to sensor malfunctions. 4. Applications to chemical processes where control is critical in achieving the desired stability and performance objectives. The research will also provide fundamental insight into the problems and limitations that sensor malfunctions cause on process control, develop practically-implementable control algorithms accounting explicitly for sensor malfunctions, address the integration of sensor fault-detection and reconfiguration methods with industrial decision support technologies, and illustrate the application of these methods to chemical processes. Broader impact. These control methods for processes subject to sensor malfunctions are expected to significantly improve the operation and performance of chemical processes, increase process safety and reliability, and minimize the negative economic impact of failures on overall process operation. This research addresses the design of feedback control and estimation systems accounting explicitly for the occurrence of sensor faults and uniquely integrates controller design, sensor fault-detection and isolation, and decision support technologies and provides the potential for significant insight on the balance that can exist between these in practical implementation. The integration of the research into education would benefit educators teaching advanced-level classes in process control and operations. The development of software, short courses and workshops, and the collaboration with the members of an industrial consortium will be the means for transferring the results of this research into the industrial sector.
